davmre

16 ms·

91.
by davmre 11y ago
Compression is a subtle and powerful thing. The ability to compress is closely related to prediction: if you can predict 95% of the moves a chess grandmaster will make, you can compress the game by explicitly representing only the other 5%
92.
by davmre 11y ago
The lines are infinitely thin. Equivalently you can imagine the paper is divided into regions of width L, and the question is whether the needle will cross a region boundary ( https://en.wikipedia.org/wiki/Buffon's_
93.
by davmre 11y ago
Andrew Ng's Coursera ML course is supposed to be pretty accessible. I've also heard good things about Machine Learning for Hackers ( http://www.amazon.com/Machine-Learning-Hackers-Drew-Conway/d... ). Ultimately
94.
by davmre 11y ago
Haskell is actually great for writing certain types of web services. You can get close to C-level speed, but with the security and correctness guarantees of strong static type checking. Obviously it has its own warts (performance optimizati
95.
by davmre 11y ago
The distributed part of the system probably depends on other components of Google's proprietary infrastructure, e.g., for cluster job scheduling. So it's a bit of work to tear that out and replace with something that's extern
96.
by davmre 11y ago
There are already a few university ML courses out there using Theano (for which Tensorflow is essentially a drop-in replacement), and I think this will be a much bigger trend over the next few years. IMHO for a first course it's useful
97.
by davmre 11y ago
The idea of representing a program by a declarative computation graph of matrix transformations has been big in deep learning research for a few years. Theano [1] is the canonical example, though more recently the space has gotten bigger wi
98.
by davmre 11y ago
Deep learning works well when you have a ton of data and don't care about interpreting the parameters of your model. My understanding is this is pretty much the opposite of the typical situation in econ research.
99.
by davmre 11y ago
I'm sure they do this to some extent, but I'd guess that prolific users are rare, and the vast majority of reviews are written by users who have only written a small number of other reviews. So the danger is that downweighting or
100.
by davmre 11y ago
The land value tax isn't the problem here. We can discourage development through zoning, environmental policy, etc. In the absence of explicit restrictions, the market value of coastal land will reflect its potential for development,
101.
by davmre 11y ago
The point is that supervised training on large networks requires millions or billions of examples. In RL, that means potentially billions of execution trajectories, which can be really expensive if you're executing on an actual, physic
102.
by davmre 11y ago
That's a silly argument. The point of the ad is that the hotel taxes allow the city to offer better public services. Of course if your analysis only considers 1.4% of the tax money you'll find the benefits aren't that impress
103.
by davmre 11y ago
There's a very reasonable distinction to be made between "gambling" bets, which are zero sum, and investments, which are not. If I invest emotional energy in a friendship, that's certainly a risk which might not always p
104.
by davmre 11y ago
Computer scientists win MacArthur grants almost every year [0]. This year it was Chris Re, a Stanford prof. Most of the CS grants are to academics, but not all, e.g., RMS got one in 1990. [0] https://en.wikipedia.org/wiki&#x
105.
by davmre 11y ago
There's essentially no evidence that women are biologically predisposed against CS. There's quite a lot of evidence that social factors do push women away. That doesn't guarantee that there's no biological component, but
106.
by davmre 11y ago
The greedy answer: because systematically discouraging 50% of the brainpower of the human species from entering one of our most productive and simultaneously short-staffed occupations is bad for technological progress and the growth of glob
107.
by davmre 11y ago
It's a joke, consistent with Matt Levine's writing style. See for example the image and caption for his article on Panda Express: http://www.bloombergview.com/articles/2015-09-24/panda-expre...
108.
by davmre 11y ago
In a healthy, well-regulated market, corporations compete to provide products and services. This provides everyone with incentives to improve quality and lower costs, and prevents a single corporation from charging monopoly rents, which is
109.
by davmre 11y ago
No, he didn't say that at all. "We’d guess that’s not what he had in mind" is Techcrunch editorializing. Musk himself didn't say what he had in mind. Given that 620 miles is roughly triple a Tesla's current everyday
110.
by davmre 11y ago
A trained network should be considered compiled code. It's almost literally a circuit wiring diagram, the lowest level of all code representations. Arguably the "source" in this case is the dataset used to train the model, al
111.
by davmre 11y ago
Locality-sensitive hashing should work: http://www.slaney.org/malcolm/yahoo/Slaney2008-LSHTutorial.p... https://en.wikipedia.org/wiki/Locality-sensitive_hashing
112.
by davmre 11y ago
My compilers course used Java to implement a simplified Java-like language. Writing a compiler in Java isn't as gloriously straightforward as in Lisp/ML/Haskell, but it's a lot less annoying than C.
113.
by davmre 11y ago
Reinforcement learning seems like overkill if you're just trying to match a single target. RL agents want to maximize reward accumulated over time, while you'd be happy to find a single good state. You could frame this as stochast
114.
by davmre 11y ago
There's a big difference between hourly and salaried positions. It's harder to define "off the clock" when no one is actually clocking you in the first place.
115.
by davmre 11y ago
> The latter is what gives the symbols meaning and humans are not threatened by the former. For what it's worth, I think you're wildly wrong a) that the Chinese room experiment has anything profound to say about conscious exper
116.
by davmre 11y ago
What I never understood: once you freeze the burritos, isn't the freshness advantage essentially moot? You might as well just ship them in trucks to the frozen-foods aisles of NYC grocery stores.
117.
by davmre 11y ago
> A Sierra Nevada Bigfoot has 30.3 [2]. That number is the total carbs, not sugar specifically. Most beers have relatively low sugar content; your link actually claims 0g of sugar for the Sierra Nevada Bigfoot. Calories in beer come part
118.
by davmre 11y ago
Sure, everyone using a laptop (or desktop!) should run regular backups. I think that's mostly orthogonal to the portability solutions being discussed here, though. For example, the VM-on-a-USB-stick solution would also require a separa
119.
by davmre 11y ago
A couple of ideas that don't exactly fit your question but might be relevant to others in a similar position: - You could buy a more portable laptop that you wouldn't mind carrying around, for example the new Macbook Retina or an
120.
by davmre 11y ago
Replacing runtime dependencies is doable for dynamically linked libraries as long as they provide the same ABI, e.g., you can swap out BLAS implementations in this way. Obviously static linking is a different story.

More ›